[cvs] / xvidcore / src / decoder.c Repository:
ViewVC logotype

Annotation of /xvidcore/src/decoder.c

Parent Directory Parent Directory | Revision Log Revision Log

Revision 1.49.2.20 - (view) (download)

1 : edgomez 1.48 /*****************************************************************************
2 : Isibaar 1.1 *
3 : edgomez 1.12 * XVID MPEG-4 VIDEO CODEC
4 : edgomez 1.48 * - Decoder Module -
5 : Isibaar 1.1 *
6 : edgomez 1.49.2.7 * Copyright(C) 2002 MinChen <chenm001@163.com>
7 :     * 2002-2003 Peter Ross <pross@xvid.org>
8 : edgomez 1.42 *
9 : edgomez 1.49.2.7 * This program is free software ; you can redistribute it and/or modify
10 : edgomez 1.47 * it under the terms of the GNU General Public License as published by
11 : edgomez 1.49.2.7 * the Free Software Foundation ; either version 2 of the License, or
12 : edgomez 1.12 * (at your option) any later version.
13 :     *
14 :     * This program is distributed in the hope that it will be useful,
15 : edgomez 1.49.2.7 * but WITHOUT ANY WARRANTY ; without even the implied warranty of
16 : edgomez 1.12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 :     * GNU General Public License for more details.
18 :     *
19 :     * You should have received a copy of the GNU General Public License
20 : edgomez 1.49.2.7 * along with this program ; if not, write to the Free Software
21 : edgomez 1.12 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22 : Isibaar 1.1 *
23 : edgomez 1.49.2.18 * $Id$
24 : edgomez 1.43 *
25 : edgomez 1.48 ****************************************************************************/
26 : Isibaar 1.1
27 : edgomez 1.47 #include <stdio.h>
28 : Isibaar 1.1 #include <stdlib.h>
29 : edgomez 1.12 #include <string.h>
30 : Isibaar 1.1
31 : chenm001 1.27 #ifdef BFRAMES_DEC_DEBUG
32 :     #define BFRAMES_DEC
33 :     #endif
34 :    
35 : Isibaar 1.1 #include "xvid.h"
36 :     #include "portab.h"
37 : edgomez 1.47 #include "global.h"
38 : Isibaar 1.1
39 :     #include "decoder.h"
40 :     #include "bitstream/bitstream.h"
41 :     #include "bitstream/mbcoding.h"
42 :    
43 : edgomez 1.49.2.15 #include "quant/quant.h"
44 : edgomez 1.49.2.20 #include "quant/quant_matrix.h"
45 : Isibaar 1.1 #include "dct/idct.h"
46 :     #include "dct/fdct.h"
47 :     #include "utils/mem_transfer.h"
48 :     #include "image/interpolate8x8.h"
49 : edgomez 1.47 #include "image/reduced.h"
50 :     #include "image/font.h"
51 : Isibaar 1.1
52 :     #include "bitstream/mbcoding.h"
53 :     #include "prediction/mbprediction.h"
54 :     #include "utils/timer.h"
55 :     #include "utils/emms.h"
56 : edgomez 1.47 #include "motion/motion.h"
57 : chl 1.49.2.9 #include "motion/gmc.h"
58 : Isibaar 1.1
59 :     #include "image/image.h"
60 :     #include "image/colorspace.h"
61 : Isibaar 1.3 #include "utils/mem_align.h"
62 : Isibaar 1.1
63 : syskin 1.49.2.17 static int
64 : edgomez 1.47 decoder_resize(DECODER * dec)
65 : Isibaar 1.1 {
66 : edgomez 1.47 /* free existing */
67 :     image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
68 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
69 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
70 :     image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
71 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
72 :    
73 :     image_destroy(&dec->gmc, dec->edged_width, dec->edged_height);
74 : Isibaar 1.1
75 : edgomez 1.49.2.13 if (dec->last_mbs)
76 : edgomez 1.47 xvid_free(dec->last_mbs);
77 :     if (dec->mbs)
78 :     xvid_free(dec->mbs);
79 : Isibaar 1.1
80 : edgomez 1.47 /* realloc */
81 : Isibaar 1.1 dec->mb_width = (dec->width + 15) / 16;
82 :     dec->mb_height = (dec->height + 15) / 16;
83 :    
84 :     dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE;
85 :     dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE;
86 : edgomez 1.19
87 :     if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) {
88 : Isibaar 1.3 xvid_free(dec);
89 : Isibaar 1.1 return XVID_ERR_MEMORY;
90 :     }
91 :    
92 : edgomez 1.19 if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) {
93 : Isibaar 1.1 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
94 : Isibaar 1.3 xvid_free(dec);
95 : Isibaar 1.1 return XVID_ERR_MEMORY;
96 :     }
97 : edgomez 1.47
98 : edgomez 1.48 /* Support B-frame to reference last 2 frame */
99 : edgomez 1.19 if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) {
100 : chenm001 1.11 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
101 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
102 :     xvid_free(dec);
103 :     return XVID_ERR_MEMORY;
104 :     }
105 : edgomez 1.47 if (image_create(&dec->tmp, dec->edged_width, dec->edged_height)) {
106 :     image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
107 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
108 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
109 :     xvid_free(dec);
110 :     return XVID_ERR_MEMORY;
111 :     }
112 :    
113 :     if (image_create(&dec->qtmp, dec->edged_width, dec->edged_height)) {
114 :     image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
115 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
116 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
117 :     image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
118 :     xvid_free(dec);
119 :     return XVID_ERR_MEMORY;
120 :     }
121 :    
122 :     if (image_create(&dec->gmc, dec->edged_width, dec->edged_height)) {
123 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
124 : chenm001 1.14 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
125 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
126 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
127 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
128 : chenm001 1.14 xvid_free(dec);
129 :     return XVID_ERR_MEMORY;
130 :     }
131 : Isibaar 1.1
132 : edgomez 1.19 dec->mbs =
133 :     xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height,
134 :     CACHE_LINE);
135 :     if (dec->mbs == NULL) {
136 : Isibaar 1.1 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
137 : chenm001 1.14 image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
138 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
139 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
140 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
141 : chenm001 1.14 xvid_free(dec);
142 :     return XVID_ERR_MEMORY;
143 :     }
144 : Isibaar 1.20 memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height);
145 :    
146 : edgomez 1.48 /* For skip MB flag */
147 : edgomez 1.19 dec->last_mbs =
148 :     xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height,
149 :     CACHE_LINE);
150 :     if (dec->last_mbs == NULL) {
151 : chenm001 1.14 xvid_free(dec->mbs);
152 :     image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
153 :     image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
154 :     image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
155 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
156 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
157 : Isibaar 1.3 xvid_free(dec);
158 : Isibaar 1.1 return XVID_ERR_MEMORY;
159 :     }
160 : Isibaar 1.20
161 :     memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height);
162 : Isibaar 1.1
163 : suxen_drol 1.49.2.1 return 0;
164 : edgomez 1.47 }
165 :    
166 :    
167 :     int
168 : suxen_drol 1.49.2.1 decoder_create(xvid_dec_create_t * create)
169 : edgomez 1.47 {
170 :     DECODER *dec;
171 :    
172 : edgomez 1.49.2.10 if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */
173 : suxen_drol 1.49.2.1 return XVID_ERR_VERSION;
174 :    
175 : edgomez 1.47 dec = xvid_malloc(sizeof(DECODER), CACHE_LINE);
176 :     if (dec == NULL) {
177 :     return XVID_ERR_MEMORY;
178 :     }
179 : edgomez 1.49.2.20
180 : edgomez 1.47 memset(dec, 0, sizeof(DECODER));
181 :    
182 : edgomez 1.49.2.20 dec->mpeg_quant_matrices = xvid_malloc(sizeof(uint16_t) * 64 * 8, CACHE_LINE);
183 :     if (dec->mpeg_quant_matrices == NULL) {
184 :     xvid_free(dec);
185 :     return XVID_ERR_MEMORY;
186 :     }
187 :    
188 : suxen_drol 1.49.2.1 create->handle = dec;
189 : edgomez 1.47
190 : suxen_drol 1.49.2.1 dec->width = create->width;
191 :     dec->height = create->height;
192 : edgomez 1.47
193 :     image_null(&dec->cur);
194 :     image_null(&dec->refn[0]);
195 :     image_null(&dec->refn[1]);
196 :     image_null(&dec->tmp);
197 :     image_null(&dec->qtmp);
198 :    
199 : edgomez 1.48 /* image based GMC */
200 : edgomez 1.47 image_null(&dec->gmc);
201 :    
202 :    
203 :     dec->mbs = NULL;
204 :     dec->last_mbs = NULL;
205 :    
206 : Isibaar 1.1 init_timer();
207 : edgomez 1.49.2.20 init_mpeg_matrix(dec->mpeg_quant_matrices);
208 : Isibaar 1.1
209 : edgomez 1.48 /* For B-frame support (used to save reference frame's time */
210 : edgomez 1.47 dec->frames = 0;
211 : chenm001 1.14 dec->time = dec->time_base = dec->last_time_base = 0;
212 : edgomez 1.47 dec->low_delay = 0;
213 :     dec->packed_mode = 0;
214 :    
215 :     dec->fixed_dimensions = (dec->width > 0 && dec->height > 0);
216 : edgomez 1.19
217 : edgomez 1.47 if (dec->fixed_dimensions)
218 :     return decoder_resize(dec);
219 :     else
220 : edgomez 1.49.2.13 return 0;
221 : Isibaar 1.1 }
222 :    
223 :    
224 : edgomez 1.19 int
225 :     decoder_destroy(DECODER * dec)
226 : Isibaar 1.1 {
227 : chenm001 1.14 xvid_free(dec->last_mbs);
228 : Isibaar 1.3 xvid_free(dec->mbs);
229 : edgomez 1.47
230 : edgomez 1.48 /* image based GMC */
231 :     image_destroy(&dec->gmc, dec->edged_width, dec->edged_height);
232 : edgomez 1.47
233 : chenm001 1.11 image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height);
234 : chenm001 1.14 image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height);
235 : edgomez 1.47 image_destroy(&dec->tmp, dec->edged_width, dec->edged_height);
236 :     image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height);
237 : Isibaar 1.1 image_destroy(&dec->cur, dec->edged_width, dec->edged_height);
238 : edgomez 1.49.2.20 xvid_free(dec->mpeg_quant_matrices);
239 : Isibaar 1.3 xvid_free(dec);
240 : Isibaar 1.1
241 :     write_timer();
242 : suxen_drol 1.49.2.1 return 0;
243 : Isibaar 1.1 }
244 :    
245 : edgomez 1.19 static const int32_t dquant_table[4] = {
246 : Isibaar 1.1 -1, -2, 1, 2
247 :     };
248 :    
249 : edgomez 1.48 /* decode an intra macroblock */
250 : syskin 1.49.2.17 static void
251 : edgomez 1.19 decoder_mbintra(DECODER * dec,
252 :     MACROBLOCK * pMB,
253 :     const uint32_t x_pos,
254 :     const uint32_t y_pos,
255 :     const uint32_t acpred_flag,
256 :     const uint32_t cbp,
257 :     Bitstream * bs,
258 :     const uint32_t quant,
259 : chenm001 1.26 const uint32_t intra_dc_threshold,
260 : edgomez 1.47 const unsigned int bound,
261 :     const int reduced_resolution)
262 : Isibaar 1.1 {
263 : edgomez 1.7
264 :     DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE);
265 : edgomez 1.19 DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE);
266 : edgomez 1.7
267 : h 1.8 uint32_t stride = dec->edged_width;
268 :     uint32_t stride2 = stride / 2;
269 :     uint32_t next_block = stride * 8;
270 : h 1.5 uint32_t i;
271 :     uint32_t iQuant = pMB->quant;
272 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
273 :    
274 : edgomez 1.47 if (reduced_resolution) {
275 :     pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5);
276 :     pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4);
277 :     pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4);
278 :     }else{
279 :     pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
280 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
281 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
282 :     }
283 : edgomez 1.7
284 : edgomez 1.48 memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */
285 : h 1.5
286 : edgomez 1.19 for (i = 0; i < 6; i++) {
287 : h 1.5 uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4);
288 : Isibaar 1.1 int16_t predictors[8];
289 :     int start_coeff;
290 :    
291 :     start_timer();
292 : edgomez 1.19 predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64],
293 : chenm001 1.26 iQuant, iDcScaler, predictors, bound);
294 : edgomez 1.19 if (!acpred_flag) {
295 : h 1.5 pMB->acpred_directions[i] = 0;
296 : Isibaar 1.1 }
297 :     stop_prediction_timer();
298 :    
299 : edgomez 1.19 if (quant < intra_dc_threshold) {
300 : Isibaar 1.1 int dc_size;
301 :     int dc_dif;
302 :    
303 : edgomez 1.19 dc_size = i < 4 ? get_dc_size_lum(bs) : get_dc_size_chrom(bs);
304 :     dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0;
305 : Isibaar 1.1
306 : edgomez 1.19 if (dc_size > 8) {
307 : edgomez 1.48 BitstreamSkip(bs, 1); /* marker */
308 : Isibaar 1.1 }
309 : edgomez 1.19
310 :     block[i * 64 + 0] = dc_dif;
311 : Isibaar 1.1 start_coeff = 1;
312 : chenm001 1.26
313 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_COEFF,"block[0] %i\n", dc_dif);
314 : edgomez 1.19 } else {
315 : Isibaar 1.1 start_coeff = 0;
316 :     }
317 :    
318 :     start_timer();
319 : edgomez 1.48 if (cbp & (1 << (5 - i))) /* coded */
320 : Isibaar 1.1 {
321 : edgomez 1.47 int direction = dec->alternate_vertical_scan ?
322 :     2 : pMB->acpred_directions[i];
323 :    
324 :     get_intra_block(bs, &block[i * 64], direction, start_coeff);
325 : Isibaar 1.1 }
326 :     stop_coding_timer();
327 :    
328 :     start_timer();
329 : edgomez 1.19 add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors);
330 : Isibaar 1.1 stop_prediction_timer();
331 :    
332 :     start_timer();
333 : edgomez 1.19 if (dec->quant_type == 0) {
334 : edgomez 1.49.2.20 dequant_h263_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices);
335 : edgomez 1.19 } else {
336 : edgomez 1.49.2.20 dequant_mpeg_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices);
337 : Isibaar 1.1 }
338 :     stop_iquant_timer();
339 :    
340 :     start_timer();
341 : edgomez 1.19 idct(&data[i * 64]);
342 : Isibaar 1.1 stop_idct_timer();
343 : edgomez 1.47
344 : h 1.5 }
345 : Isibaar 1.1
346 : edgomez 1.19 if (dec->interlacing && pMB->field_dct) {
347 : h 1.8 next_block = stride;
348 :     stride *= 2;
349 : Isibaar 1.1 }
350 : h 1.5
351 :     start_timer();
352 : edgomez 1.47
353 :     if (reduced_resolution)
354 :     {
355 :     next_block*=2;
356 :     copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride);
357 :     copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride);
358 :     copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride);
359 :     copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride);
360 :     copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2);
361 :     copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2);
362 :     }else{
363 :     transfer_16to8copy(pY_Cur, &data[0 * 64], stride);
364 :     transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride);
365 :     transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride);
366 :     transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride);
367 :     transfer_16to8copy(pU_Cur, &data[4 * 64], stride2);
368 :     transfer_16to8copy(pV_Cur, &data[5 * 64], stride2);
369 :     }
370 : h 1.5 stop_transfer_timer();
371 : Isibaar 1.1 }
372 :    
373 : syskin 1.49.2.17 static void
374 :     decoder_mb_decode(DECODER * dec,
375 :     const uint32_t cbp,
376 :     Bitstream * bs,
377 :     uint8_t * pY_Cur,
378 :     uint8_t * pU_Cur,
379 :     uint8_t * pV_Cur,
380 :     const int reduced_resolution,
381 :     const MACROBLOCK * pMB)
382 :     {
383 :     DECLARE_ALIGNED_MATRIX(block, 1, 64, int16_t, CACHE_LINE);
384 :     DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE);
385 : Isibaar 1.1
386 : edgomez 1.49.2.18 int stride = dec->edged_width;
387 :     int next_block = stride * (reduced_resolution ? 16 : 8);
388 : syskin 1.49.2.17 const int stride2 = stride/2;
389 :     int i;
390 :     const uint32_t iQuant = pMB->quant;
391 :     const int direction = dec->alternate_vertical_scan ? 2 : 0;
392 :     const quant_interFuncPtr dequant = dec->quant_type == 0 ? dequant_h263_inter : dequant_mpeg_inter;
393 : Isibaar 1.1
394 : syskin 1.49.2.17 for (i = 0; i < 6; i++) {
395 :    
396 :     if (cbp & (1 << (5 - i))) { /* coded */
397 :    
398 :     memset(block, 0, 64 * sizeof(int16_t)); /* clear */
399 :    
400 :     start_timer();
401 :     get_inter_block(bs, block, direction);
402 :     stop_coding_timer();
403 :    
404 :     start_timer();
405 : edgomez 1.49.2.20 dequant(&data[i * 64], block, iQuant, dec->mpeg_quant_matrices);
406 : syskin 1.49.2.17 stop_iquant_timer();
407 :    
408 :     start_timer();
409 :     idct(&data[i * 64]);
410 :     stop_idct_timer();
411 :     }
412 :     }
413 :    
414 :     if (dec->interlacing && pMB->field_dct) {
415 :     next_block = stride;
416 :     stride *= 2;
417 :     }
418 :    
419 :     start_timer();
420 :     if (reduced_resolution) {
421 :     if (cbp & 32)
422 :     add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride);
423 :     if (cbp & 16)
424 :     add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride);
425 :     if (cbp & 8)
426 :     add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride);
427 :     if (cbp & 4)
428 :     add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride);
429 :     if (cbp & 2)
430 :     add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2);
431 :     if (cbp & 1)
432 :     add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2);
433 :     } else {
434 :     if (cbp & 32)
435 :     transfer_16to8add(pY_Cur, &data[0 * 64], stride);
436 :     if (cbp & 16)
437 :     transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride);
438 :     if (cbp & 8)
439 :     transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride);
440 :     if (cbp & 4)
441 :     transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride);
442 :     if (cbp & 2)
443 :     transfer_16to8add(pU_Cur, &data[4 * 64], stride2);
444 :     if (cbp & 1)
445 :     transfer_16to8add(pV_Cur, &data[5 * 64], stride2);
446 :     }
447 :     stop_transfer_timer();
448 :     }
449 : Isibaar 1.1
450 : edgomez 1.48 /* decode an inter macroblock */
451 : syskin 1.49.2.17 static void
452 : edgomez 1.19 decoder_mbinter(DECODER * dec,
453 :     const MACROBLOCK * pMB,
454 :     const uint32_t x_pos,
455 :     const uint32_t y_pos,
456 :     const uint32_t cbp,
457 :     Bitstream * bs,
458 : edgomez 1.47 const uint32_t rounding,
459 : syskin 1.49.2.17 const int reduced_resolution,
460 :     const int ref)
461 : Isibaar 1.1 {
462 : h 1.8 uint32_t stride = dec->edged_width;
463 :     uint32_t stride2 = stride / 2;
464 : edgomez 1.7 uint32_t i;
465 : syskin 1.49.2.17
466 : h 1.5 uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
467 : edgomez 1.47
468 : Isibaar 1.1 int uv_dx, uv_dy;
469 : edgomez 1.47 VECTOR mv[4]; /* local copy of mvs */
470 :    
471 :     if (reduced_resolution) {
472 :     pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5);
473 :     pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4);
474 :     pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4);
475 :     for (i = 0; i < 4; i++) {
476 :     mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x);
477 :     mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y);
478 :     }
479 :     } else {
480 :     pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
481 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
482 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
483 :     for (i = 0; i < 4; i++)
484 :     mv[i] = pMB->mvs[i];
485 :     }
486 : edgomez 1.49.2.13
487 : syskin 1.49.2.17 start_timer();
488 : edgomez 1.47
489 : syskin 1.49.2.17 if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */
490 : edgomez 1.49.2.13
491 : syskin 1.49.2.17 uv_dx = mv[0].x;
492 :     uv_dy = mv[0].y;
493 :     if (dec->quarterpel) {
494 :     uv_dx /= 2;
495 :     uv_dy /= 2;
496 :     }
497 : edgomez 1.47 uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
498 :     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
499 :    
500 :     if (reduced_resolution)
501 :     interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos,
502 : syskin 1.49.2.17 mv[0].x, mv[0].y, stride, rounding);
503 :     else if (dec->quarterpel)
504 :     interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64,
505 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
506 : edgomez 1.47 mv[0].x, mv[0].y, stride, rounding);
507 :     else
508 : syskin 1.49.2.17 interpolate16x16_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos,
509 :     mv[0].x, mv[0].y, stride, rounding);
510 : edgomez 1.47
511 : syskin 1.49.2.17 } else { /* MODE_INTER4V */
512 : edgomez 1.47
513 : syskin 1.49.2.17 if(dec->quarterpel) {
514 :     uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2);
515 :     uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2);
516 :     } else {
517 :     uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x;
518 :     uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y;
519 :     }
520 : edgomez 1.47
521 : syskin 1.49.2.17 uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
522 :     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
523 : edgomez 1.47
524 : syskin 1.49.2.17 if (reduced_resolution) {
525 : edgomez 1.47 interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos,
526 : syskin 1.49.2.17 mv[0].x, mv[0].y, stride, rounding);
527 : edgomez 1.47 interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos,
528 : syskin 1.49.2.17 mv[1].x, mv[1].y, stride, rounding);
529 : edgomez 1.47 interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16,
530 : syskin 1.49.2.17 mv[2].x, mv[2].y, stride, rounding);
531 : edgomez 1.49.2.13 interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16,
532 : syskin 1.49.2.17 mv[3].x, mv[3].y, stride, rounding);
533 : edgomez 1.47 interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos,
534 : syskin 1.49.2.17 uv_dx, uv_dy, stride2, rounding);
535 : edgomez 1.47 interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos,
536 : syskin 1.49.2.17 uv_dx, uv_dy, stride2, rounding);
537 : edgomez 1.47
538 : syskin 1.49.2.17 } else if (dec->quarterpel) {
539 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
540 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
541 :     mv[0].x, mv[0].y, stride, rounding);
542 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
543 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
544 :     mv[1].x, mv[1].y, stride, rounding);
545 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
546 :     dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
547 :     mv[2].x, mv[2].y, stride, rounding);
548 :     interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64,
549 :     dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
550 :     mv[3].x, mv[3].y, stride, rounding);
551 :     } else {
552 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos,
553 :     mv[0].x, mv[0].y, stride, rounding);
554 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos,
555 :     mv[1].x, mv[1].y, stride, rounding);
556 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos + 8,
557 :     mv[2].x, mv[2].y, stride, rounding);
558 :     interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos + 8,
559 :     mv[3].x, mv[3].y, stride, rounding);
560 : edgomez 1.47 }
561 :     }
562 :    
563 : syskin 1.49.2.17 /* chroma */
564 :     if (reduced_resolution) {
565 :     interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos,
566 :     uv_dx, uv_dy, stride2, rounding);
567 :     interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos,
568 :     uv_dx, uv_dy, stride2, rounding);
569 :     } else {
570 :     interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos,
571 :     uv_dx, uv_dy, stride2, rounding);
572 :     interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos,
573 :     uv_dx, uv_dy, stride2, rounding);
574 : edgomez 1.47 }
575 : Isibaar 1.1
576 : syskin 1.49.2.17 stop_comp_timer();
577 : Isibaar 1.34
578 : syskin 1.49.2.17 if (cbp)
579 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur,
580 :     reduced_resolution, pMB);
581 : edgomez 1.47 }
582 :    
583 :     static void
584 :     decoder_mbgmc(DECODER * dec,
585 :     MACROBLOCK * const pMB,
586 :     const uint32_t x_pos,
587 :     const uint32_t y_pos,
588 :     const uint32_t fcode,
589 :     const uint32_t cbp,
590 :     Bitstream * bs,
591 : syskin 1.49.2.17 const uint32_t rounding)
592 : edgomez 1.47 {
593 :     const uint32_t stride = dec->edged_width;
594 :     const uint32_t stride2 = stride / 2;
595 : syskin 1.49.2.17
596 : edgomez 1.47 uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
597 :     uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
598 :     uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
599 :    
600 : syskin 1.49.2.17 NEW_GMC_DATA * gmc_data = &dec->new_gmc_data;
601 :    
602 : edgomez 1.47 pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
603 : Isibaar 1.1
604 :     start_timer();
605 : edgomez 1.49.2.13
606 : edgomez 1.47 /* this is where the calculations are done */
607 : edgomez 1.49.2.13
608 : syskin 1.49.2.17 gmc_data->predict_16x16(gmc_data,
609 :     dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y,
610 :     stride, stride, x_pos, y_pos, rounding);
611 :    
612 :     gmc_data->predict_8x8(gmc_data,
613 :     dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u,
614 :     dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v,
615 :     stride2, stride2, x_pos, y_pos, rounding);
616 : chl 1.49.2.9
617 : syskin 1.49.2.17 gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel);
618 : chl 1.49.2.9
619 : syskin 1.49.2.17 pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode);
620 :     pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode);
621 : chl 1.49.2.9
622 : edgomez 1.47 pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv;
623 : edgomez 1.49.2.13
624 : edgomez 1.47 stop_transfer_timer();
625 : edgomez 1.49.2.13
626 : syskin 1.49.2.17 if (cbp)
627 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB);
628 : Isibaar 1.1
629 :     }
630 :    
631 :    
632 : syskin 1.49.2.17 static void
633 : edgomez 1.19 decoder_iframe(DECODER * dec,
634 : syskin 1.49.2.17 Bitstream * bs,
635 :     int reduced_resolution,
636 :     int quant,
637 :     int intra_dc_threshold)
638 : Isibaar 1.1 {
639 : chenm001 1.26 uint32_t bound;
640 :     uint32_t x, y;
641 : edgomez 1.47 uint32_t mb_width = dec->mb_width;
642 :     uint32_t mb_height = dec->mb_height;
643 : edgomez 1.49.2.13
644 : syskin 1.49.2.17 if (reduced_resolution) {
645 : edgomez 1.47 mb_width = (dec->width + 31) / 32;
646 :     mb_height = (dec->height + 31) / 32;
647 :     }
648 : chenm001 1.25
649 : chenm001 1.26 bound = 0;
650 : edgomez 1.7
651 : edgomez 1.47 for (y = 0; y < mb_height; y++) {
652 :     for (x = 0; x < mb_width; x++) {
653 : chenm001 1.26 MACROBLOCK *mb;
654 : Isibaar 1.1 uint32_t mcbpc;
655 :     uint32_t cbpc;
656 :     uint32_t acpred_flag;
657 :     uint32_t cbpy;
658 :     uint32_t cbp;
659 :    
660 : chenm001 1.26 while (BitstreamShowBits(bs, 9) == 1)
661 :     BitstreamSkip(bs, 9);
662 :    
663 :     if (check_resync_marker(bs, 0))
664 :     {
665 : edgomez 1.49.2.13 bound = read_video_packet_header(bs, dec, 0,
666 : edgomez 1.47 &quant, NULL, NULL, &intra_dc_threshold);
667 :     x = bound % mb_width;
668 :     y = bound / mb_width;
669 : chenm001 1.26 }
670 :     mb = &dec->mbs[y * dec->mb_width + x];
671 :    
672 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
673 : chenm001 1.26
674 : Isibaar 1.1 mcbpc = get_mcbpc_intra(bs);
675 :     mb->mode = mcbpc & 7;
676 :     cbpc = (mcbpc >> 4);
677 :    
678 :     acpred_flag = BitstreamGetBit(bs);
679 :    
680 :     cbpy = get_cbpy(bs, 1);
681 :     cbp = (cbpy << 2) | cbpc;
682 :    
683 : edgomez 1.19 if (mb->mode == MODE_INTRA_Q) {
684 :     quant += dquant_table[BitstreamGetBits(bs, 2)];
685 :     if (quant > 31) {
686 : Isibaar 1.1 quant = 31;
687 : edgomez 1.19 } else if (quant < 1) {
688 : Isibaar 1.1 quant = 1;
689 :     }
690 :     }
691 :     mb->quant = quant;
692 : chenm001 1.30 mb->mvs[0].x = mb->mvs[0].y =
693 :     mb->mvs[1].x = mb->mvs[1].y =
694 :     mb->mvs[2].x = mb->mvs[2].y =
695 :     mb->mvs[3].x = mb->mvs[3].y =0;
696 : h 1.5
697 : edgomez 1.19 if (dec->interlacing) {
698 : h 1.5 mb->field_dct = BitstreamGetBit(bs);
699 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct);
700 : h 1.5 }
701 : Isibaar 1.1
702 : edgomez 1.19 decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
703 : edgomez 1.47 intra_dc_threshold, bound, reduced_resolution);
704 :    
705 : Isibaar 1.1 }
706 : albeu 1.32 if(dec->out_frm)
707 : syskin 1.49.2.17 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width);
708 : Isibaar 1.1 }
709 : edgomez 1.7
710 : Isibaar 1.1 }
711 :    
712 :    
713 : syskin 1.49.2.17 static void
714 : edgomez 1.19 get_motion_vector(DECODER * dec,
715 : syskin 1.49.2.17 Bitstream * bs,
716 :     int x,
717 :     int y,
718 :     int k,
719 :     VECTOR * ret_mv,
720 :     int fcode,
721 :     const int bound)
722 : Isibaar 1.1 {
723 : edgomez 1.7
724 : syskin 1.49.2.17 const int scale_fac = 1 << (fcode - 1);
725 :     const int high = (32 * scale_fac) - 1;
726 :     const int low = ((-32) * scale_fac);
727 :     const int range = (64 * scale_fac);
728 : Isibaar 1.1
729 : syskin 1.49.2.17 const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k);
730 : edgomez 1.47 VECTOR mv;
731 : Isibaar 1.1
732 : edgomez 1.47 mv.x = get_mv(bs, fcode);
733 :     mv.y = get_mv(bs, fcode);
734 : edgomez 1.19
735 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MV,"mv_diff (%i,%i) pred (%i,%i) result (%i,%i)\n", mv.x, mv.y, pmv.x, pmv.y, mv.x+pmv.x, mv.y+pmv.y);
736 : chenm001 1.26
737 : edgomez 1.47 mv.x += pmv.x;
738 :     mv.y += pmv.y;
739 : Isibaar 1.1
740 : edgomez 1.47 if (mv.x < low) {
741 :     mv.x += range;
742 :     } else if (mv.x > high) {
743 :     mv.x -= range;
744 : Isibaar 1.1 }
745 :    
746 : edgomez 1.47 if (mv.y < low) {
747 :     mv.y += range;
748 :     } else if (mv.y > high) {
749 :     mv.y -= range;
750 : Isibaar 1.1 }
751 :    
752 : edgomez 1.47 ret_mv->x = mv.x;
753 :     ret_mv->y = mv.y;
754 :     }
755 :    
756 :     /* for P_VOP set gmc_warp to NULL */
757 : syskin 1.49.2.17 static void
758 : edgomez 1.19 decoder_pframe(DECODER * dec,
759 : syskin 1.49.2.17 Bitstream * bs,
760 :     int rounding,
761 :     int reduced_resolution,
762 :     int quant,
763 :     int fcode,
764 :     int intra_dc_threshold,
765 :     const WARPPOINTS *const gmc_warp)
766 : Isibaar 1.1 {
767 :     uint32_t x, y;
768 : chenm001 1.26 uint32_t bound;
769 : albeu 1.32 int cp_mb, st_mb;
770 : edgomez 1.47 uint32_t mb_width = dec->mb_width;
771 :     uint32_t mb_height = dec->mb_height;
772 : edgomez 1.49.2.13
773 : syskin 1.49.2.17 if (reduced_resolution) {
774 : edgomez 1.47 mb_width = (dec->width + 31) / 32;
775 :     mb_height = (dec->height + 31) / 32;
776 :     }
777 : Isibaar 1.1
778 :     start_timer();
779 : edgomez 1.19 image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
780 : syskin 1.49.2.17 dec->width, dec->height);
781 : Isibaar 1.1 stop_edges_timer();
782 :    
783 : syskin 1.49.2.17 if (gmc_warp) {
784 :     /* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */
785 : edgomez 1.49.2.13 generate_GMCparameters( dec->sprite_warping_points,
786 :     dec->sprite_warping_accuracy, gmc_warp,
787 : chl 1.49.2.9 dec->width, dec->height, &dec->new_gmc_data);
788 : edgomez 1.47
789 : syskin 1.49.2.17 /* image warping is done block-based in decoder_mbgmc(), now */
790 : edgomez 1.47 }
791 :    
792 : chenm001 1.26 bound = 0;
793 :    
794 : edgomez 1.47 for (y = 0; y < mb_height; y++) {
795 : albeu 1.32 cp_mb = st_mb = 0;
796 : edgomez 1.47 for (x = 0; x < mb_width; x++) {
797 : chenm001 1.26 MACROBLOCK *mb;
798 :    
799 : edgomez 1.48 /* skip stuffing */
800 : chenm001 1.26 while (BitstreamShowBits(bs, 10) == 1)
801 :     BitstreamSkip(bs, 10);
802 :    
803 : syskin 1.49.2.17 if (check_resync_marker(bs, fcode - 1)) {
804 : edgomez 1.49.2.13 bound = read_video_packet_header(bs, dec, fcode - 1,
805 : edgomez 1.47 &quant, &fcode, NULL, &intra_dc_threshold);
806 :     x = bound % mb_width;
807 :     y = bound / mb_width;
808 : chenm001 1.26 }
809 :     mb = &dec->mbs[y * dec->mb_width + x];
810 :    
811 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32));
812 : suxen_drol 1.23
813 : syskin 1.49.2.17 if (!(BitstreamGetBit(bs))) { /* block _is_ coded */
814 :     uint32_t mcbpc, cbpc, cbpy, cbp;
815 :     uint32_t intra, acpred_flag = 0;
816 : edgomez 1.48 int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */
817 : Isibaar 1.1
818 : albeu 1.32 cp_mb++;
819 : Isibaar 1.1 mcbpc = get_mcbpc_inter(bs);
820 :     mb->mode = mcbpc & 7;
821 :     cbpc = (mcbpc >> 4);
822 : edgomez 1.49.2.13
823 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode);
824 :     DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc);
825 : Isibaar 1.1
826 :     intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q);
827 : edgomez 1.19
828 : edgomez 1.47 if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q))
829 :     mcsel = BitstreamGetBit(bs);
830 : syskin 1.49.2.17 else if (intra)
831 : edgomez 1.49.2.12 acpred_flag = BitstreamGetBit(bs);
832 : edgomez 1.47
833 : chenm001 1.26 cbpy = get_cbpy(bs, intra);
834 : syskin 1.49.2.17 DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel);
835 : chenm001 1.25
836 : Isibaar 1.1 cbp = (cbpy << 2) | cbpc;
837 :    
838 : edgomez 1.19 if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) {
839 : chenm001 1.26 int dquant = dquant_table[BitstreamGetBits(bs, 2)];
840 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant);
841 : chenm001 1.26 quant += dquant;
842 : edgomez 1.19 if (quant > 31) {
843 : Isibaar 1.1 quant = 31;
844 : chenm001 1.26 } else if (quant < 1) {
845 : Isibaar 1.1 quant = 1;
846 :     }
847 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant);
848 : Isibaar 1.1 }
849 :     mb->quant = quant;
850 : h 1.5
851 : edgomez 1.19 if (dec->interlacing) {
852 : syskin 1.49.2.14 if ((cbp || intra) && !mcsel) {
853 : h 1.37 mb->field_dct = BitstreamGetBit(bs);
854 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
855 : h 1.37 }
856 : h 1.5
857 : edgomez 1.19 if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) {
858 : h 1.5 mb->field_pred = BitstreamGetBit(bs);
859 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
860 : h 1.5
861 : edgomez 1.19 if (mb->field_pred) {
862 : h 1.5 mb->field_for_top = BitstreamGetBit(bs);
863 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
864 : h 1.5 mb->field_for_bot = BitstreamGetBit(bs);
865 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
866 : h 1.5 }
867 :     }
868 :     }
869 : edgomez 1.49.2.13
870 : edgomez 1.47 if (mcsel) {
871 : syskin 1.49.2.17 decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding);
872 : edgomez 1.47 continue;
873 :    
874 :     } else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) {
875 : h 1.5
876 : edgomez 1.19 if (dec->interlacing && mb->field_pred) {
877 : syskin 1.49.2.17 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
878 :     get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode, bound);
879 : edgomez 1.19 } else {
880 : syskin 1.49.2.17 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
881 : edgomez 1.47 mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
882 : h 1.5 }
883 : chl 1.35 } else if (mb->mode == MODE_INTER4V ) {
884 : chenm001 1.26 get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound);
885 :     get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound);
886 :     get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound);
887 :     get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound);
888 : syskin 1.49.2.17 } else { /* MODE_INTRA, MODE_INTRA_Q */
889 :     mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
890 :     mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
891 : edgomez 1.19 decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant,
892 : edgomez 1.47 intra_dc_threshold, bound, reduced_resolution);
893 : Isibaar 1.1 continue;
894 :     }
895 :    
896 : syskin 1.49.2.17 decoder_mbinter(dec, mb, x, y, cbp, bs,
897 :     rounding, reduced_resolution, 0);
898 : edgomez 1.47
899 : syskin 1.49.2.17 } else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */
900 : edgomez 1.47 mb->mode = MODE_NOT_CODED_GMC;
901 : syskin 1.49.2.17 decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, rounding);
902 : edgomez 1.39
903 : edgomez 1.47 if(dec->out_frm && cp_mb > 0) {
904 : syskin 1.49.2.17 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
905 :     cp_mb = 0;
906 : edgomez 1.47 }
907 :     st_mb = x+1;
908 : syskin 1.49.2.17 } else { /* not coded P_VOP macroblock */
909 : chenm001 1.14 mb->mode = MODE_NOT_CODED;
910 : edgomez 1.47
911 : Isibaar 1.1 mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0;
912 :     mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0;
913 : edgomez 1.49.2.13
914 : syskin 1.49.2.17 decoder_mbinter(dec, mb, x, y, 0, bs,
915 :     rounding, reduced_resolution, 0);
916 : edgomez 1.47
917 : albeu 1.32 if(dec->out_frm && cp_mb > 0) {
918 : syskin 1.49.2.17 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
919 :     cp_mb = 0;
920 : albeu 1.32 }
921 :     st_mb = x+1;
922 : Isibaar 1.1 }
923 :     }
924 : syskin 1.49.2.17
925 : albeu 1.32 if(dec->out_frm && cp_mb > 0)
926 : syskin 1.49.2.17 output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb);
927 : Isibaar 1.1 }
928 :     }
929 :    
930 : edgomez 1.47
931 : edgomez 1.48 /* decode B-frame motion vector */
932 : syskin 1.49.2.17 static void
933 :     get_b_motion_vector(Bitstream * bs,
934 : edgomez 1.47 VECTOR * mv,
935 :     int fcode,
936 :     const VECTOR pmv)
937 : chenm001 1.14 {
938 : syskin 1.49.2.17 const int scale_fac = 1 << (fcode - 1);
939 :     const int high = (32 * scale_fac) - 1;
940 :     const int low = ((-32) * scale_fac);
941 :     const int range = (64 * scale_fac);
942 : edgomez 1.7
943 : syskin 1.49.2.17 int mv_x = get_mv(bs, fcode);
944 :     int mv_y = get_mv(bs, fcode);
945 : Isibaar 1.1
946 : syskin 1.49.2.17 mv_x += pmv.x;
947 :     mv_y += pmv.y;
948 : albeu 1.32
949 : syskin 1.49.2.17 if (mv_x < low)
950 : edgomez 1.47 mv_x += range;
951 : syskin 1.49.2.17 else if (mv_x > high)
952 : edgomez 1.47 mv_x -= range;
953 :    
954 : syskin 1.49.2.17 if (mv_y < low)
955 : edgomez 1.47 mv_y += range;
956 : syskin 1.49.2.17 else if (mv_y > high)
957 : edgomez 1.47 mv_y -= range;
958 :    
959 :     mv->x = mv_x;
960 :     mv->y = mv_y;
961 :     }
962 :    
963 : syskin 1.49.2.16 /* decode an B-frame direct & interpolate macroblock */
964 : syskin 1.49.2.17 static void
965 : edgomez 1.47 decoder_bf_interpolate_mbinter(DECODER * dec,
966 : syskin 1.49.2.17 IMAGE forward,
967 :     IMAGE backward,
968 :     const MACROBLOCK * pMB,
969 :     const uint32_t x_pos,
970 :     const uint32_t y_pos,
971 :     Bitstream * bs,
972 :     const int direct)
973 : edgomez 1.47 {
974 :     uint32_t stride = dec->edged_width;
975 :     uint32_t stride2 = stride / 2;
976 :     int uv_dx, uv_dy;
977 :     int b_uv_dx, b_uv_dy;
978 :     uint8_t *pY_Cur, *pU_Cur, *pV_Cur;
979 : syskin 1.49.2.17 const uint32_t cbp = pMB->cbp;
980 : edgomez 1.47
981 :     pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4);
982 :     pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3);
983 :     pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3);
984 :    
985 : syskin 1.49.2.16 if (!direct) {
986 : edgomez 1.47 uv_dx = pMB->mvs[0].x;
987 :     uv_dy = pMB->mvs[0].y;
988 :    
989 :     b_uv_dx = pMB->b_mvs[0].x;
990 :     b_uv_dy = pMB->b_mvs[0].y;
991 :    
992 : syskin 1.49.2.17 if (dec->quarterpel) {
993 : edgomez 1.47 uv_dx /= 2;
994 :     uv_dy /= 2;
995 :     b_uv_dx /= 2;
996 :     b_uv_dy /= 2;
997 :     }
998 :    
999 :     uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3];
1000 :     uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3];
1001 :    
1002 :     b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3];
1003 :     b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3];
1004 :    
1005 : syskin 1.49.2.17 } else {
1006 :     if(dec->quarterpel) {
1007 :     uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2);
1008 :     uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2);
1009 :     b_uv_dx = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2);
1010 :     b_uv_dy = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2);
1011 :     } else {
1012 :     uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x;
1013 :     uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y;
1014 :     b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x;
1015 :     b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y;
1016 :     }
1017 : edgomez 1.47
1018 : syskin 1.49.2.17 uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf];
1019 :     uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf];
1020 :     b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf];
1021 :     b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf];
1022 : edgomez 1.47 }
1023 :    
1024 :     start_timer();
1025 :     if(dec->quarterpel) {
1026 : syskin 1.49.2.17 if(!direct) {
1027 : edgomez 1.47 interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1028 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1029 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1030 :     } else {
1031 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1032 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1033 :     pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1034 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1035 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
1036 :     pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
1037 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1038 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
1039 :     pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
1040 : edgomez 1.47 interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64,
1041 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
1042 :     pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
1043 : edgomez 1.47 }
1044 : syskin 1.49.2.17 } else {
1045 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos,
1046 : syskin 1.49.2.17 pMB->mvs[0].x, pMB->mvs[0].y, stride, 0);
1047 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos,
1048 : syskin 1.49.2.17 pMB->mvs[1].x, pMB->mvs[1].y, stride, 0);
1049 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8,
1050 : syskin 1.49.2.17 pMB->mvs[2].x, pMB->mvs[2].y, stride, 0);
1051 : edgomez 1.47 interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8,
1052 : syskin 1.49.2.17 16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, 0);
1053 : edgomez 1.47 }
1054 :    
1055 :     interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx,
1056 : syskin 1.49.2.17 uv_dy, stride2, 0);
1057 : edgomez 1.47 interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx,
1058 : syskin 1.49.2.17 uv_dy, stride2, 0);
1059 : edgomez 1.47
1060 :    
1061 :     if(dec->quarterpel) {
1062 : syskin 1.49.2.17 if(!direct) {
1063 : edgomez 1.47 interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1064 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1065 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1066 :     } else {
1067 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1068 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos, 16*y_pos,
1069 :     pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1070 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1071 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos,
1072 :     pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
1073 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1074 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8,
1075 :     pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
1076 : edgomez 1.47 interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64,
1077 : syskin 1.49.2.17 dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8,
1078 :     pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
1079 : edgomez 1.47 }
1080 : syskin 1.49.2.17 } else {
1081 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos,
1082 : syskin 1.49.2.17 pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0);
1083 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8,
1084 : syskin 1.49.2.17 16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0);
1085 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos,
1086 : syskin 1.49.2.17 16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0);
1087 : edgomez 1.47 interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8,
1088 : syskin 1.49.2.17 16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0);
1089 : edgomez 1.47 }
1090 :    
1091 :     interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos,
1092 : syskin 1.49.2.17 b_uv_dx, b_uv_dy, stride2, 0);
1093 : edgomez 1.47 interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos,
1094 : syskin 1.49.2.17 b_uv_dx, b_uv_dy, stride2, 0);
1095 : edgomez 1.47
1096 :     interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
1097 :     dec->cur.y + (16 * y_pos * stride) + 16 * x_pos,
1098 :     dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos,
1099 :     stride, 1, 8);
1100 :    
1101 :     interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8,
1102 :     dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8,
1103 :     dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8,
1104 :     stride, 1, 8);
1105 :    
1106 :     interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
1107 :     dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
1108 :     dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos,
1109 :     stride, 1, 8);
1110 :    
1111 :     interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
1112 :     dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
1113 :     dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8,
1114 :     stride, 1, 8);
1115 :    
1116 :     interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos,
1117 :     dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos,
1118 :     dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos,
1119 :     stride2, 1, 8);
1120 :    
1121 :     interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos,
1122 :     dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos,
1123 :     dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos,
1124 :     stride2, 1, 8);
1125 :    
1126 :     stop_comp_timer();
1127 :    
1128 : syskin 1.49.2.17 if (cbp)
1129 :     decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB);
1130 : edgomez 1.47 }
1131 :    
1132 : edgomez 1.48 /* for decode B-frame dbquant */
1133 : syskin 1.49.2.17 static __inline int32_t
1134 : edgomez 1.47 get_dbquant(Bitstream * bs)
1135 :     {
1136 : syskin 1.49.2.17 if (!BitstreamGetBit(bs)) /* '0' */
1137 : edgomez 1.47 return (0);
1138 : syskin 1.49.2.17 else if (!BitstreamGetBit(bs)) /* '10' */
1139 : edgomez 1.47 return (-2);
1140 : syskin 1.49.2.17 else /* '11' */
1141 : edgomez 1.48 return (2);
1142 : edgomez 1.47 }
1143 :    
1144 : edgomez 1.48 /*
1145 : syskin 1.49.2.17 * decode B-frame mb_type
1146 :     * bit ret_value
1147 :     * 1 0
1148 :     * 01 1
1149 :     * 001 2
1150 :     * 0001 3
1151 : edgomez 1.48 */
1152 : syskin 1.49.2.17 static int32_t __inline
1153 : edgomez 1.47 get_mbtype(Bitstream * bs)
1154 :     {
1155 :     int32_t mb_type;
1156 :    
1157 : syskin 1.49.2.17 for (mb_type = 0; mb_type <= 3; mb_type++)
1158 : edgomez 1.47 if (BitstreamGetBit(bs))
1159 : syskin 1.49.2.17 return (mb_type);
1160 : edgomez 1.47
1161 : syskin 1.49.2.17 return -1;
1162 : edgomez 1.47 }
1163 :    
1164 : syskin 1.49.2.17 static void
1165 : edgomez 1.47 decoder_bframe(DECODER * dec,
1166 : syskin 1.49.2.17 Bitstream * bs,
1167 :     int quant,
1168 :     int fcode_forward,
1169 :     int fcode_backward)
1170 : edgomez 1.47 {
1171 :     uint32_t x, y;
1172 :     VECTOR mv;
1173 :     const VECTOR zeromv = {0,0};
1174 : syskin 1.49.2.17 const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp;
1175 :     int i;
1176 :    
1177 : edgomez 1.47 start_timer();
1178 :     image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height,
1179 : syskin 1.49.2.17 dec->width, dec->height);
1180 : edgomez 1.47 image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height,
1181 : syskin 1.49.2.17 dec->width, dec->height);
1182 : edgomez 1.47 stop_edges_timer();
1183 :    
1184 :     for (y = 0; y < dec->mb_height; y++) {
1185 : edgomez 1.48 /* Initialize Pred Motion Vector */
1186 : edgomez 1.47 dec->p_fmv = dec->p_bmv = zeromv;
1187 :     for (x = 0; x < dec->mb_width; x++) {
1188 :     MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x];
1189 :     MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x];
1190 : edgomez 1.49.2.19 const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward;
1191 :     uint32_t intra_dc_threshold; /* fake variable */
1192 :    
1193 :     if (check_resync_marker(bs, fcode_max - 1)) {
1194 :     int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant,
1195 :     &fcode_forward, &fcode_backward, &intra_dc_threshold);
1196 :     x = bound % dec->mb_width;
1197 :     y = bound / dec->mb_width;
1198 :     /* reset predicted macroblocks */
1199 :     dec->p_fmv = dec->p_bmv = zeromv;
1200 :     }
1201 : edgomez 1.47
1202 :     mv =
1203 :     mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] =
1204 :     mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv;
1205 : syskin 1.49.2.17 mb->quant = quant;
1206 : edgomez 1.47
1207 : edgomez 1.48 /*
1208 : edgomez 1.49.2.13 * skip if the co-located P_VOP macroblock is not coded
1209 : edgomez 1.48 * if not codec in co-located S_VOP macroblock is _not_
1210 :     * automatically skipped
1211 :     */
1212 : edgomez 1.47
1213 :     if (last_mb->mode == MODE_NOT_CODED) {
1214 :     mb->cbp = 0;
1215 : syskin 1.49.2.16 mb->mode = MODE_FORWARD;
1216 : syskin 1.49.2.17 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1);
1217 : edgomez 1.47 continue;
1218 :     }
1219 :    
1220 : edgomez 1.48 if (!BitstreamGetBit(bs)) { /* modb=='0' */
1221 : edgomez 1.47 const uint8_t modb2 = BitstreamGetBit(bs);
1222 :    
1223 : syskin 1.49.2.16 mb->mode = get_mbtype(bs);
1224 : edgomez 1.47
1225 : syskin 1.49.2.17 if (!modb2) /* modb=='00' */
1226 : edgomez 1.47 mb->cbp = BitstreamGetBits(bs, 6);
1227 : syskin 1.49.2.17 else
1228 : edgomez 1.47 mb->cbp = 0;
1229 : syskin 1.49.2.17
1230 : syskin 1.49.2.16 if (mb->mode && mb->cbp) {
1231 : edgomez 1.47 quant += get_dbquant(bs);
1232 : syskin 1.49.2.17 if (quant > 31)
1233 : edgomez 1.47 quant = 31;
1234 : syskin 1.49.2.17 else if (quant < 1)
1235 : edgomez 1.47 quant = 1;
1236 :     }
1237 : syskin 1.49.2.17 mb->quant = quant;
1238 : syskin 1.49.2.14
1239 :     if (dec->interlacing) {
1240 :     if (mb->cbp) {
1241 :     mb->field_dct = BitstreamGetBit(bs);
1242 :     DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct);
1243 :     }
1244 :    
1245 : syskin 1.49.2.16 if (mb->mode) {
1246 : syskin 1.49.2.14 mb->field_pred = BitstreamGetBit(bs);
1247 :     DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred);
1248 :    
1249 :     if (mb->field_pred) {
1250 :     mb->field_for_top = BitstreamGetBit(bs);
1251 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top);
1252 :     mb->field_for_bot = BitstreamGetBit(bs);
1253 :     DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot);
1254 :     }
1255 :     }
1256 :     }
1257 :    
1258 : edgomez 1.47 } else {
1259 : syskin 1.49.2.16 mb->mode = MODE_DIRECT_NONE_MV;
1260 : edgomez 1.47 mb->cbp = 0;
1261 :     }
1262 :    
1263 : syskin 1.49.2.16 switch (mb->mode) {
1264 : edgomez 1.47 case MODE_DIRECT:
1265 : syskin 1.49.2.17 get_b_motion_vector(bs, &mv, 1, zeromv);
1266 : edgomez 1.47
1267 :     case MODE_DIRECT_NONE_MV:
1268 : syskin 1.49.2.17 for (i = 0; i < 4; i++) {
1269 :     mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) / TRD + mv.x);
1270 :     mb->b_mvs[i].x = (int32_t) ((mv.x == 0)
1271 :     ? ((TRB - TRD) * last_mb->mvs[i].x) / TRD
1272 :     : mb->mvs[i].x - last_mb->mvs[i].x);
1273 :     mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) / TRD + mv.y);
1274 :     mb->b_mvs[i].y = (int32_t) ((mv.y == 0)
1275 :     ? ((TRB - TRD) * last_mb->mvs[i].y) / TRD
1276 :     : mb->mvs[i].y - last_mb->mvs[i].y);
1277 : edgomez 1.47 }
1278 : syskin 1.49.2.17
1279 : edgomez 1.47 decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
1280 : syskin 1.49.2.17 mb, x, y, bs, 1);
1281 : edgomez 1.47 break;
1282 :    
1283 :     case MODE_INTERPOLATE:
1284 : syskin 1.49.2.17 get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv);
1285 : edgomez 1.47 dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1286 :    
1287 : syskin 1.49.2.17 get_b_motion_vector(bs, &mb->b_mvs[0], fcode_backward, dec->p_bmv);
1288 :     dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = mb->b_mvs[0];
1289 : edgomez 1.47
1290 :     decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0],
1291 : syskin 1.49.2.17 mb, x, y, bs, 0);
1292 : edgomez 1.47 break;
1293 :    
1294 :     case MODE_BACKWARD:
1295 : syskin 1.49.2.17 get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv);
1296 : edgomez 1.47 dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1297 :    
1298 : syskin 1.49.2.17 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 0);
1299 : edgomez 1.47 break;
1300 :    
1301 :     case MODE_FORWARD:
1302 : syskin 1.49.2.17 get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv);
1303 : edgomez 1.47 dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0];
1304 :    
1305 : syskin 1.49.2.17 decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1);
1306 : edgomez 1.47 break;
1307 :    
1308 :     default:
1309 : syskin 1.49.2.16 DPRINTF(XVID_DEBUG_ERROR,"Not supported B-frame mb_type = %i\n", mb->mode);
1310 : edgomez 1.47 }
1311 : edgomez 1.48 } /* End of for */
1312 :     }
1313 : edgomez 1.47 }
1314 :    
1315 :     /* perform post processing if necessary, and output the image */
1316 : edgomez 1.49.2.13 void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs,
1317 : suxen_drol 1.49.2.1 xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, int coding_type)
1318 : edgomez 1.47 {
1319 :     image_output(img, dec->width, dec->height,
1320 : suxen_drol 1.49.2.1 dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride,
1321 :     frame->output.csp, dec->interlacing);
1322 :    
1323 : syskin 1.49.2.17 if (stats) {
1324 : suxen_drol 1.49.2.1 stats->type = coding2type(coding_type);
1325 :     stats->data.vop.time_base = (int)dec->time_base;
1326 : edgomez 1.49.2.6 stats->data.vop.time_increment = 0; /* XXX: todo */
1327 : suxen_drol 1.49.2.1 }
1328 : edgomez 1.47 }
1329 :    
1330 :    
1331 :     int
1332 :     decoder_decode(DECODER * dec,
1333 : syskin 1.49.2.17 xvid_dec_frame_t * frame, xvid_dec_stats_t * stats)
1334 : edgomez 1.47 {
1335 :    
1336 :     Bitstream bs;
1337 :     uint32_t rounding;
1338 :     uint32_t reduced_resolution;
1339 :     uint32_t quant;
1340 :     uint32_t fcode_forward;
1341 :     uint32_t fcode_backward;
1342 :     uint32_t intra_dc_threshold;
1343 :     WARPPOINTS gmc_warp;
1344 : suxen_drol 1.49.2.1 int coding_type;
1345 :     int success, output, seen_something;
1346 : edgomez 1.47
1347 : edgomez 1.49.2.10 if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */
1348 : suxen_drol 1.49.2.1 return XVID_ERR_VERSION;
1349 : edgomez 1.47
1350 : suxen_drol 1.49.2.1 start_global_timer();
1351 : edgomez 1.47
1352 : suxen_drol 1.49.2.1 dec->low_delay_default = (frame->general & XVID_LOWDELAY);
1353 :     if ((frame->general & XVID_DISCONTINUITY))
1354 : edgomez 1.47 dec->frames = 0;
1355 : suxen_drol 1.49.2.1 dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL;
1356 : edgomez 1.47
1357 : syskin 1.49.2.17 if (frame->length < 0) { /* decoder flush */
1358 :     int ret;
1359 :     /* if not decoding "low_delay/packed", and this isn't low_delay and
1360 :     we have a reference frame, then outout the reference frame */
1361 : suxen_drol 1.49.2.3 if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) {
1362 : suxen_drol 1.49.2.1 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type);
1363 : syskin 1.49.2.17 dec->frames = 0;
1364 :     ret = 0;
1365 :     } else {
1366 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1367 :     ret = XVID_ERR_END;
1368 :     }
1369 : edgomez 1.47
1370 :     emms();
1371 :     stop_global_timer();
1372 : suxen_drol 1.49.2.3 return ret;
1373 : edgomez 1.47 }
1374 : edgomez 1.19
1375 : Isibaar 1.1 BitstreamInit(&bs, frame->bitstream, frame->length);
1376 :    
1377 : edgomez 1.48 /* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */
1378 : edgomez 1.47 if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f)
1379 :     {
1380 :     image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width,
1381 : suxen_drol 1.49.2.1 (uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing);
1382 :     if (stats) stats->type = XVID_TYPE_NOTHING;
1383 : edgomez 1.47 emms();
1384 : syskin 1.49.2.17 return 1; /* one byte consumed */
1385 : edgomez 1.47 }
1386 :    
1387 : suxen_drol 1.49.2.1 success = 0;
1388 :     output = 0;
1389 :     seen_something = 0;
1390 :    
1391 : edgomez 1.47 repeat:
1392 :    
1393 : syskin 1.49.2.17 coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution,
1394 : edgomez 1.47 &quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp);
1395 :    
1396 : edgomez 1.49.2.13 DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i\n",
1397 : suxen_drol 1.49.2.1 coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp);
1398 : Isibaar 1.1
1399 : syskin 1.49.2.17 if (coding_type == -1) { /* nothing */
1400 : edgomez 1.47 if (success) goto done;
1401 : syskin 1.49.2.17 if (stats) stats->type = XVID_TYPE_NOTHING;
1402 : edgomez 1.47 emms();
1403 : syskin 1.49.2.17 return BitstreamPos(&bs)/8;
1404 : Isibaar 1.1 }
1405 :    
1406 : syskin 1.49.2.17 if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */
1407 :    
1408 : suxen_drol 1.49.2.1 if (coding_type == -3)
1409 : edgomez 1.47 decoder_resize(dec);
1410 : edgomez 1.49.2.13
1411 : syskin 1.49.2.17 if (stats) {
1412 : suxen_drol 1.49.2.1 stats->type = XVID_TYPE_VOL;
1413 : edgomez 1.47 stats->data.vol.general = 0;
1414 : suxen_drol 1.49.2.1 /*XXX: if (dec->interlacing)
1415 :     stats->data.vol.general |= ++INTERLACING; */
1416 : edgomez 1.47 stats->data.vol.width = dec->width;
1417 :     stats->data.vol.height = dec->height;
1418 : suxen_drol 1.49.2.1 stats->data.vol.par = dec->aspect_ratio;
1419 : edgomez 1.47 stats->data.vol.par_width = dec->par_width;
1420 :     stats->data.vol.par_height = dec->par_height;
1421 :     emms();
1422 : suxen_drol 1.49.2.1 return BitstreamPos(&bs)/8; /* number of bytes consumed */
1423 : edgomez 1.47 }
1424 :     goto repeat;
1425 : edgomez 1.49.2.13 }
1426 : edgomez 1.47
1427 : edgomez 1.48 dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */
1428 : edgomez 1.47
1429 :     /* packed_mode: special-N_VOP treament */
1430 : syskin 1.49.2.17 if (dec->packed_mode && coding_type == N_VOP) {
1431 :     if (dec->low_delay_default && dec->frames > 0) {
1432 : suxen_drol 1.49.2.1 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type);
1433 : edgomez 1.47 output = 1;
1434 :     }
1435 :     /* ignore otherwise */
1436 : syskin 1.49.2.17 } else if (coding_type != B_VOP) {
1437 :     switch(coding_type) {
1438 : edgomez 1.47 case I_VOP :
1439 :     decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold);
1440 :     break;
1441 :     case P_VOP :
1442 : edgomez 1.49.2.13 decoder_pframe(dec, &bs, rounding, reduced_resolution, quant,
1443 : edgomez 1.47 fcode_forward, intra_dc_threshold, NULL);
1444 :     break;
1445 :     case S_VOP :
1446 : edgomez 1.49.2.13 decoder_pframe(dec, &bs, rounding, reduced_resolution, quant,
1447 : edgomez 1.47 fcode_forward, intra_dc_threshold, &gmc_warp);
1448 :     break;
1449 :     case N_VOP :
1450 : edgomez 1.49.2.6 /* XXX: not_coded vops are not used for forward prediction */
1451 :     /* we should not swap(last_mbs,mbs) */
1452 : edgomez 1.47 image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height);
1453 :     break;
1454 :     }
1455 :    
1456 : syskin 1.49.2.17 if (reduced_resolution) {
1457 : edgomez 1.47 image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs,
1458 :     (dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width,
1459 : suxen_drol 1.49.2.1 16, 0);
1460 : edgomez 1.47 }
1461 : Isibaar 1.1
1462 : edgomez 1.47 /* note: for packed_mode, output is performed when the special-N_VOP is decoded */
1463 : syskin 1.49.2.17 if (!(dec->low_delay_default && dec->packed_mode)) {
1464 :     if (dec->low_delay) {
1465 : suxen_drol 1.49.2.1 decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type);
1466 : edgomez 1.47 output = 1;
1467 : syskin 1.49.2.17 } else if (dec->frames > 0) { /* is the reference frame valid? */
1468 : edgomez 1.47 /* output the reference frame */
1469 : suxen_drol 1.49.2.1 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type);
1470 : edgomez 1.47 output = 1;
1471 :     }
1472 :     }
1473 : Isibaar 1.18
1474 : chenm001 1.14 image_swap(&dec->refn[0], &dec->refn[1]);
1475 :     image_swap(&dec->cur, &dec->refn[0]);
1476 : syskin 1.49.2.17 SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs);
1477 : edgomez 1.47 dec->last_reduced_resolution = reduced_resolution;
1478 : syskin 1.49.2.17 dec->last_coding_type = coding_type;
1479 : edgomez 1.47
1480 :     dec->frames++;
1481 :     seen_something = 1;
1482 : chenm001 1.31
1483 : syskin 1.49.2.17 } else { /* B_VOP */
1484 : edgomez 1.47
1485 : syskin 1.49.2.17 if (dec->low_delay) {
1486 : edgomez 1.49.2.5 DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n");
1487 : edgomez 1.47 dec->low_delay = 1;
1488 :     }
1489 :    
1490 : syskin 1.49.2.17 if (dec->frames < 2) {
1491 : edgomez 1.47 /* attemping to decode a bvop without atleast 2 reference frames */
1492 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1493 :     "broken b-frame, mising ref frames");
1494 : syskin 1.49.2.17 } else if (dec->time_pp <= dec->time_bp) {
1495 : edgomez 1.49.2.13 /* this occurs when dx50_bvop_compatibility==0 sequences are
1496 : edgomez 1.47 decoded in vfw. */
1497 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1498 :     "broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp);
1499 : syskin 1.49.2.17 } else {
1500 : edgomez 1.47 decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward);
1501 :     }
1502 :    
1503 : suxen_drol 1.49.2.1 decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type);
1504 : edgomez 1.47 output = 1;
1505 :     dec->frames++;
1506 : chenm001 1.14 }
1507 :    
1508 : edgomez 1.47 BitstreamByteAlign(&bs);
1509 :    
1510 :     /* low_delay_default mode: repeat in packed_mode */
1511 : syskin 1.49.2.17 if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) {
1512 : edgomez 1.47 success = 1;
1513 :     goto repeat;
1514 :     }
1515 :    
1516 :     done :
1517 :    
1518 :     /* low_delay_default mode: if we've gotten here without outputting anything,
1519 :     then output the recently decoded frame, or print an error message */
1520 : syskin 1.49.2.17 if (dec->low_delay_default && output == 0) {
1521 :     if (dec->packed_mode && seen_something) {
1522 : edgomez 1.47 /* output the recently decoded frame */
1523 : suxen_drol 1.49.2.1 decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type);
1524 : syskin 1.49.2.17 } else {
1525 : edgomez 1.47 image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128);
1526 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16,
1527 :     "warning: nothing to output");
1528 :     image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64,
1529 :     "bframe decoder lag");
1530 :    
1531 : edgomez 1.49.2.13 decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP);
1532 : suxen_drol 1.49.2.1 if (stats) stats->type = XVID_TYPE_NOTHING;
1533 : edgomez 1.47 }
1534 :     }
1535 :    
1536 : Isibaar 1.1 emms();
1537 :     stop_global_timer();
1538 :    
1539 : suxen_drol 1.49.2.1 return BitstreamPos(&bs) / 8; /* number of bytes consumed */
1540 : Isibaar 1.1 }
No admin address has been configured
 ViewVC Help
Powered by ViewVC 1.0.4